edge2vec: Representation learning using edge semantics for biomedical knowledge discovery

BMC Bioinformatics

Table 3 Compound-gene bioactivity prediction

Algorithm	Precision	Recall	F1 measure	Hamming loss	AUROC
DeepWalk	0.7787	0.7750	0.7742	0.2250	0.7660
LINE	0.8170	0.8166	0.8166	0.1833	0.8058
node2vec	0.7983	0.7916	0.7904	0.2083	0.7793
metapath2vec (Co-Ge-Co)	0.5170	0.5170	0.5168	0.4830	0.5007
metapath2vec (Co-Ge-Ge-Co)	0.4979	0.4980	0.4976	0.5020	0.4890
metapath2vec (Co-Dr-Ge-Dr-Co)	0.5305	0.5305	0.5304	0.4695	0.5304
metapath2vec++ (Co-Ge-Co)	0.4969	0.4970	0.4965	0.5030	0.4776
metapath2vec++ (Co-Ge-Ge-Co)	0.4854	0.4855	0.4854	0.5145	0.4776
metapath2vec++ (Co-Dr-Ge-Dr-Co)	0.5120	0.5120	0.5119	0.4880	0.5102
edge2vec	0.9017*	0.9000*	0.8998*	0.1000*	0.8914*

Symbol “*” highlights the cases where our model significantly beats the best baseline with p value smaller than 0.01

ISSN: 1471-2105